Scientists have confirmed that the tens of thousands of earthquakes that rattled the Greek island of Santorini earlier this year were caused by molten rock pushing through a deep underground channel for nearly three months.

More than 25,000 tremors travelled roughly 20km horizontally through the Earth’s crust. Researchers used the quakes themselves as “virtual sensors,” applying physics-based modelling and artificial intelligence to trace how the seismic activity evolved.

Dr Stephen Hicks of UCL said this blend of physics and machine learning could significantly improve the forecasting of volcanic eruptions.

What unfolded beneath Santorini?

Starting in January 2025, seismic unrest intensified beneath Santorini, Amorgos, and Anafi, with thousands of quakes above magnitude 5.0 felt across the islands. Tourists left in large numbers, while residents feared an impending eruption from the nearby underwater volcano Kolumbo—or even a repeat of the catastrophic 1956 magnitude-7.7 quake.

Publishing in Science, the research team built a 3D model showing how magma moved sideways through a 30km-long channel, over 10km below the seafloor, fed by systems beneath both Santorini and Kolumbo.
The volume of magma travelling underground was so large it could fill 200,000 Olympic pools, breaking through rock layers and triggering the months-long quake swarm.

Lead author Anthony Lomax said the pattern of earthquakes aligned precisely with what would be expected from horizontal magma intrusion.

Is the unrest now over?

According to the researchers, the episode has ended. The magma stayed deep—below 8km—and appears to have eventually cooled and solidified. Although magma can rise quickly to the surface, the sharp drop in activity suggests no imminent eruption.

Volcanoes, however, can remain unstable for years. Recent Icelandic activity demonstrates how long-lived such unrest can be.
The team believes AI-driven analysis, paired with core geophysical principles, could transform global volcano monitoring and help predict dangerous changes earlier.

“When we see clusters of quakes,” Dr Hicks said, “that data can reveal what is most likely happening underground.”